The Dry Valleys

Bing maps aerial imagery

And on the ground

Yup! There’s streetview imagery

Building a hiking network

Start with a regular lattice across the study area

Many topologies could be used

We went with hexagonal at ~100m spacing

Up \(\neq\) down

Slope is critical to movement, so attach heights to nodes and calculate slope in each direction along lattice edges

Then apply a hiking function to estimate traversal times

Hiking functions

Smoothing GPS data

Curve fitting to cleaned data

Terrain-dependent vs combined

Completing and applying the network

Assign edges estimated traversal times based on slope, the hiking function, and land cover (moraine or rock)

Make the nodes and edges into a directed graph

Then use various graph algorithms to find, e.g., everywhere-to-everywhere shortest paths

Betweenness centrality

After some exploration, we settled on betweenness centrality as a graph metric that might be useful

This measure counts how many times each node appears on the shortest paths between every other pair of nodes

An indicator of the relative likelihood of each location being visited

Radius-limited betweenness

Restrict betweenness centrality to nodes no more than some time apart

This is much faster to calculate (yay!)

Also… it looks like it has more value

It (perhaps) is relevant to how people navigate in such environments

Planning a path network?

One way to minimise impact might be to plan paths

This is experimental at this stage

Based on a minimum spanning tree approximation to an arborescence

Conclusions

Terrain-differentiated data-fitted hiking functions are a novelty

Potential wider application of radius-limited betweenness centrality?

A manuscript almost finished

For more

• On this project: dosull.github.io/antarctica

• About Geospatial Stuff: dosull.github.io